coderetreat-china / coderetreat-china.github.io

:kimono: Coderetreat 中国, work in progress for https://github.com/coderetreat/GDCR/pull/12
http://coderetreat-china.github.io/
MIT License
2 stars 1 forks source link

Readings #1

Open paulvi opened 9 years ago

paulvi commented 9 years ago

a. Architect role http://www.ibm.com/developerworks/rational/library/mar06/eeles/index.html

paulvi commented 9 years ago

Let’s look at basic Agile sprint. All team follow Agile, but not all for example

  1. collect number to draw burndown chart (overhead about 0.5-1 hour per sprint)
  2. do real velocity calculation, e.g. how much is estimation unit in man_days (overhead about 0.5-1 hour per sprint)
  3. capacity calculation (translating estimation units into man_days, can we really finish those stories put into sprint)

Without numbers, it is impossible to apply math.

  1. But if they are, there is again additional overhead about 0.5 hour per sprint (e.g. to put those number in Excel sheet and get some projections )

Agile is saying to be good as team can and improve over time, but there’s no such thing as Level 1, Level 2 of Agile

Now if we say to a team, “look, if you do all that numbers work, you are actually doing QWM - Quantitative Work Management that is part of Level 4 of CMMI, cool ah?. Team member don’t need to learn of CMMI, but it is just a line or grade that our team may set as (long term) goal.”

So the question “do we need to spent 1 hour per sprint to stay on Level 4” is much easier to answer than “how much time and work we need to spend for planning activities”. It is also more tradable with client managers. (Similar to showing some static code analysis results vs repeating “we need more time to improve architecture, software internal quality”) *1

Now how we can get managers support for such initiative and still keep developers happy (not overburdened ) with how they choose to work.

For me there are more questions, than answers though.

1: Economics of Software Quality http://ardalis.com/economics-of-software-quality